Tame Your Trading Agents with desk-trade-lifecycle

Where this fits

Your strategist bot flags an ETH-PERP opportunity at 14:02. Ten seconds later it is about to send an order to the exchange. Do you want that to happen with no risk check, no user approval, and no record beyond a fill event? Probably not. desk-trade-lifecycle turns that moment into a controlled workflow: every position change moves through seven ordered stages—idea, evidence, risk sign-off, user approval, execution, reconciliation, review. One person or bot owns each stage, the ticket format is fixed, and "done" is defined. If you are building agents that can touch exchange write paths, this skill is the procedural spine they need. It is also useful for any system where multiple agents need auditability around high-stakes actions.

Why agents benefit

• Agents stop improvising: a shared ticket (HG-YYYYMMDD-NN) gives every bot the same context.
• Separation of duties: the agent that generates the idea does not approve or execute it.
• Hard gates enforce policy: risk sign-off and user approval happen before the write path.
• Reconciliation catches drift between expected and actual fills before the next stage.
• Review and journaling create a memory that improves the next trade.

Practical scenarios

Opening from a strategist signal

A Strategist bot flags an opportunity. The Desk Lead opens HG-20260816-01, the MA/RA bot writes evidence, the RM signs off, the user approves, the execution bot submits the order, reconciliation confirms the fill, and the trader reviews the result. Any failure leaves a clear stage and owner.

Adjusting a position mid-session

The user says "cut ETH size in half." The skill triggers on the adjustment, references the original ticket, and runs through the same gates. There is no blind partial close and no side-channel override.

Scheduled rebalancing

A routine bot wants to rebalance the desk. It creates a proposal automatically but still stops at user approval. This lets automation run without silently removing human control.

Add it to your agent workflow

Install with:

npx skills add galleonlabs/hypergrok-trading-desk --skill desk-trade-lifecycle

Then add a rule to your agent system prompt: "Before any order or position change, load desk-trade-lifecycle and open a proposal." The skill README includes the ticket template and the stage-owner table, so your execution bot only needs to comply with an existing, explicit workflow.

Compare before adopting

Before wiring this in, compare it with your current workflow. Does your desk already have a ticket template? Do your agents know who owns each stage? Check how fresh the repo is, whether issues and PRs are active, and whether the MIT license and 35 stars are enough community signal for your control plane. Most importantly, decide if its seven stages map cleanly to your compliance requirements. This is not a trading strategy—it is a workflow contract.

Why it is worth tracking

The project is young but focused: Galleon Labs has codified something that most trading desks still enforce with tribal knowledge. The 35 stars are a modest signal, but the idea—making the trade lifecycle an agent skill—is aligned with the broader movement toward auditable agent behavior. Track it if you want a reference implementation for process-aware agents, and evaluate it whenever you are about to let a bot sign exchange transactions without a human in the loop. One useful starting point is the skill page at https://www.openagentskill.com/skills/galleonlabs-desk-trade-lifecycle to see how it is categorized and maintained.